In part two of this three-part interview with SearchDataBackup, backup expert W. Curtis Preston, founder of Truth in IT and backupcentral.com, discusses how tape is being used today and virtual server backups.
We've been hearing for years that tape backup use is falling. Is there still a place in the data center for tape backup?
W. Curtis Preston: Yes, and no. Is there a place for tape? Absolutely. Is there a place for tape backup? I have been saying for years that proper backup and recovery design does not involve, and should not involve directly, backing up across a network to a tape drive, end of story. I lost a lot of friends in the tape business because of that.
Then the question is, well, can tape play a role at all in that day-to-day backup and recovery? And the answer is yes. Once we have the data across the network, and we've finished our incremental backup. Now we need to copy the data to another location for the purpose of disaster recovery. That could be done via replication to another typically duplicated storage system or it could be done by the old way, which is copying it to tapes and handing it to a dude in a truck.
That is still the most economical way. It is absolutely not the best way, but in terms of cost, it's still much cheaper to make tape and hand it to a dude in a truck than to replicate them. I could talk for 20 minutes about the cost of that approach. There are people that make that cost argument, that replicated disks are cheaper than tape, but I don't see how they make that argument.
Tape is better at holding on to data for longer periods of time than disk is. If you archive a file on a spinning disk, you have to move it every five years, or it will corrupt. It's a little-known fact. It would get what we call bit rot. Tape does not have that issue to that degree. It's magnetic, it still has that issue, but it's measured much more. That's why you hear tape-storage time at 20 and 30 years.
The problem you have to worry about tape is device obsolescence, not bit rot. That you have to make sure, if you've got an LT0-2 tape, you've got a drive that can read it. That's only because you can separate the tape and the media, or the media and the drive, whereas in disk drives they're married, and so it's never a problem. So you just have to make sure you handle that. I've written about this and spoken about this quite a bit, that device obsolescence is a boogie man. That it's not as big of a problem as people make it out to be.
I make that point by going to eBay and searching on every tape drive that I've used for the last 20 years, and they're all sold en masse. There are hundreds and hundreds of every tape drive, including the QIC 80, which holds 80 megabytes of data. That was a piece of junk 20 years ago; it's still a piece of junk, but you can still get one. I'm sure there are also several companies that will repair those drives for you. I don't think device obsolescence is as scary as people make it out to be.
A lot of organizations now operate in environments where the majority of the servers are virtualized. Has the move to virtualized environments affected traditional backup strategies?
Preston: Significantly. It has really turned it on its head. You have two choices with a virtualized environment. Well, you have three, but the third one is really a bad idea, which is to just continue what you've always been doing. You can put backup software in your VMs, and you can run them. It's incredibly inefficient -- you've got 30, 40, 50 VMs; they all think they have their own SCSI bus, and they don't. So that's a recipe for disaster -- you don't want to do that.
A better option is to move the backup out of the VM and to do it at the hypervisor level. That way you can handle the VMs as VMs and know that they're files and not machines. You can do incremental backups against the VMDK file, block-level incremental against that file while it's being paused using VSS, etc. That way you can back up a machine which just happens to be pretending to be 50 machines.
The other way which is really interesting, which has not received as much play, is to switch to a backup software product that is VM-friendly -- such as CDP, near CDP, or just a block-level incremental backup product that duplicates from the very beginning.
What they all have in common is that they don't do periodic full backups -- they just back up the blocks that have changed since the initial full backup. And they do it throughout the day, so it's I/O friendly.
What are your top three tips for managing backups in a virtual server environment?
Preston: Well, I'd say first is that you need to realize that if you're doing it the old way, you're wasting money. You're costing your company money, which is actually a good thing, because it's the one area where people can make a cost argument. The second is to examine what the product that you're already using, what they do in that space. You're already using it, an enterprise-level backup product. Examine and test what their answer is to the solution.
The third depends on what happens with the second. The third is to examine all the other alternatives. Just realize that if you go down that third road that you are going to be making a major change.